ArticlePDF Available

Dynamic properties of the advertisement calls of gray tree frogs: Patterns of variability and female choice

Authors:

Abstract and Figures

We assessed the potential for several acoustic properties of the advertisement calls of male gray tree frogs to affect relative mating success by relating patterns of variation in these properties to minimum differences required to elicit female choice. Dynamic properties (pulse number, PN; call rate, CR; and duty cycle, DC, the ratio of call duration to call period) varied much more within bouts of calling than a static property (dominant frequency, DF) but nevertheless exhibited significant between male variation in three of four breeding seasons. Many multiply recorded males consistently produced calls with values substantially above or below mean values of males recorded on the same nights. Nightly ranges of variation in PN and CR were often greater than the minimum differences required to elicit female preferences in the laboratory. In most experiments, females chose high-PN or fast-CR calls over low-PN or slow-CR alternatives, respectively, even if the preferred stimuli were farther away or 6-10 dB lower in sound pressure level (SPL), provided that differences in PN or CR were 100%. Consistent with these results, females did not always choose the closer of two calling males in the field. Nightly ranges of variation in DF rarely equaled the minimum difference required to elicit SPL independent preferences. Females preferred a stimulus of high-PN and slow-CR over an alternative of low-PN or fast-CR with the same acoustic on-time; in two experiments, females chose calls of high-PN over low-PN alternatives even though the playback of the high-PN call was interrupted and the low-PN call was broadcast continuously. Thus, female preferences were not merely based on the total time of acoustic stimulation. Responses of females tested twice in the same experiment suggest that phenotypic variation in preference was limited in our study populations.
Content may be subject to copyright.
Behavioral Ecology Vol.
7 No. 1: 7-18
Dynamic properties
of the
advertisement calls
of gray tree frogs: patterns
of
variability
and
female choice
H. Carl Gerhardt,* Miranda
L.
Dyson,b
and
Steven
D.
Tanner*
"Division
of
Biological Sciences, University
of
Missouri, Columbia, Columbia,
MO
65211,
USA, and
•"Department
of
Psychology, University
of
Durban, Durban, South Africa
We assessed
the
potential
for
several acoustic properties
of the
advertisement calls
of
male gray tree frogs
to
affect relative
mating success
by
relating patterns
of
variation
in
these properties
to
minimum differences required
to
elicit female choice.
Dynamic properties (pulse number,
PN;
call rate,
CR; and
duty cycle, DC,
the
ratio
of
call duration
to
call period) varied much
more within bouts
of
calling than
a
static property (dominant frequency,
DF) but
nevertheless exhibited significant between-
male variation
in
three
of
four breeding seasons. Many multiply recorded males consistently produced calls with values substan-
tially above
or
below mean values
of
males recorded
on the
same nights. Nightly ranges
of
variation
in PN and CR
were often
greater than
the
minimum differences required
to
elicit female preferences
in the
laboratory.
In
most experiments, females
chose high-PN
or
fast-CR calls over low-PN
or
slow-CR alternatives, respectively, even
if the
preferred stimuli were farther away
or
6-10
dB lower
in
sound pressure level (SPL), provided that differences
in PN or
CR were 100%. Consistent with these results,
females
did not
always choose
the
closer
of
two calling males
in the
field. Nightly ranges
of
variation
in DF
rarely equaled
the
minimum difference required
to
elicit SPL^independent preferences. Females preferred
a
stimulus
of
high-PN
and
slow-CR over
an alternative
of
low-PN
or
fast-CR with
the
same acoustic on-time;
in
two experiments, females chose calls
of
high-PN over
low-
PN alternatives even though
the
playback
of the
high-PN call
was
interrupted
and the
low-PN call
was
broadcast continuously.
Thus,
female preferences were
not
merely based
on the
total time
of
acoustic stimulation. Responses
of
females tested twice
in
the same experiment suggest that phenotypic variation
in
preference
was
limited
in our
study populations. Key
words:
adver-
tisement
calls,
Amphibia, Anura, distance-independent preference, dynamic acoustic properties, female choice, Hyla
chrysoscelis,
Hyla
versicolor,
individual variation
in
preference, repeatability, sexual selection, tree frog.
[Behav Ecol
7:7-18
(1996)]
P
males
of
many species base mating decisions
on
extrava-
gant displays produced
by
males (review
in
Ryan
and
Keddy-Hector, 1992). Empirical
and
theoretical studies
of fe-
male choice
are
numerous,
and
most working hypotheses
about
the
evolution
of
female preferences fall into
one of
two
categories (Andersson, 1994). Direct selection
on
female
pref-
erences occurs when females
or
their offspring obtain
non-
genetic benefits from their choice
of
males,
e.g.,
increased
fertilization success
and
parental care
or
reduced assessment
costs;
indirect selection
on
female preferences occurs when
offspring inherit attributes
of the
chosen male that increase
their viability, sexual attractiveness,
or
both.
With
the
exception
of
selection
for
minimizing assessment
costs,
all of
these mechanisms require variability
in
displays.
If among-male differences
in
displays are inconsistent over
the
course
of a
breeding season, then
the
potential
for an
evolu-
tionary response
in the
display
and for
indirect sexual selec-
tion on preferences
is
probably limited.
If
little among-male
variability occurs
in
displays over
the
time span that individual
females choose mates, then
the
potential
for
direct
or
indirect
benefits
of
mate choice
is
limited.
In
such cases, females
should
use
mate-choice criteria that minimize search costs
and
predation risks (e.g., Arak, 1988; Searcy
and
Andersson,
1986)
and would
be
likely
to
mate with
the
first male encountered.
If females frequently bypass
the
first male they encounter,
then females,
on
average, must gain some direct
or
indirect
M.
L.
Dyson
is now at the
Biology Department, Open University,
Milton Keynes,
UK.
Received
28
October
1994;
revision
26
January
1995;
second revi-
sion
23
February
1995;
accepted
23
February
1995.
1045-2249/96/$5.00
O 1996
International Society
for
Behavioral Ecology
benefits that compensate
for the
assessment costs they incur.
For example,
in
two species
of
anurans
in
which females were
observed by-passing nearby males, female choice resulted
in
increased fertilization success (Robertson, 1990; Ryan, 1985).
Two important tasks
of
empirical studies
are to
identify
cri-
teria
by
which females choose mates independently
of
prox-
imity
and to
explore
the
extent
to
which these criteria
cor-
relate with potential benefits
of
choice.
We provide data that bear
on
these questions
in two
species
of gray tree frogs. Males form dense breeding aggregations
(choruses), females freely initiate matings with calling males,
and males provide
no
resources other than gametes (Fellers,
1979;
Gerhardt, 1991).
One
species, Hyla
chrysoscelis,
is a dip-
loid species that
has
given rise
to a
tetraploid species,
H.
ver-
sicolor,
at
least three times (Ptacek
et
al., 1994). Males produce
advertisement calls with species-specific pulse-repetition rates
(PR)
but
similar spectral properties such
as
dominant
fre-
quency
(DF)
(Gerhardt
and
Doherty, 1988); both properties
are highly stereotyped within
a
bout
of
calling (Gerhardt,
1991).
In
contrast
to
these static properties,
the
number
of
pulses
per
call (pulse number,
PN) and the
call-rate
(CR) are
highly variable, dynamic properties (Gerhardt, 1991), whose
values overlap broadly between
the two
species. Males often
change
PN and
CR during
a
bout
of
calling;
joint variation
in
PN
and CR is
affected strongly
by
chorus density (Runkle
et
al.,
1994; Wells
and
Taigen, 1986).
The
basic structure
of ad-
vertisement calls
and
patterns
of
calling
are
similar between
populations
of
H.
versicolor in
Connecticut
and
Missouri even
though these populations represent
two
different lineages
(e.g., Gerhardt
and
Ptacek, unpublished data; Klump
and
Gerhardt, 1987; Ptacek
et
al., 1994; Wells
and
Taigen, 1986).
Our first goal was
to
assess patterns
of
phenotypic variation
in
the
advertisement calls
of H.
versicolor
from
central
Mis-
by guest on July 6, 2011beheco.oxfordjournals.orgDownloaded from
Behavioral Ecology Vol. 7 No. 1
souri during four seasons. In particular, we examined die ex-
tent to which several acoustic properties varied within and
among males during a season and on a nighdy basis. These
data suggest which properties are likely to show evolutionary
responses to female choice assuming diat repeatabilities re-
flect heritability.
A second goal was to provide evidence concerning die ex-
tent to which females of H.
chrysoscelis
and H.
versicolor
are
likely to move extra distances to mate widi males with partic-
ular values of dynamic properties. Approaches to calling males
are difficult to observe because females are secretive (Sullivan
and Hinshaw, 1992), but
we
did observe several females choos-
ing die more distant of two calling males. Such extra move-
ments probably subject females to increased risk of predation
by visually orienting predators. We conducted female choice
experiments to estimate die minimum difference in PN diat
reliably elicits distance-independent preferences. Relating diis
and other estimates for odier call properties from die litera-
ture to our data on nighdy call variability, we assess die like-
lihood diat differences in dynamic and static properties are
sufficient for female choice.
Our diird goal was to explore die extent to which prefer-
ences for high values of dynamic properties can be explained
simply by differences between stimuli in die quantity of acous-
tic stimulation as measured by sound pressure level (SPL), die
ratio of sound to silence (duty cycle, DC) or bodi. We cor-
roborated and extended earlier findings reported by Klump
and Gerhardt (1987)' concerning preferences for long, slow-
CR calls having die same DC as short, fast-CR alternatives. We
also assessed die extent to which females move extra distances
to return to interrupted playbacks of high-PN calls in die face
of continuous stimulation by low-PN calls of higher SPL. Fi-
nally, we report die results of preliminary experiments diat
assessed phenotypic variation in preference for different val-
ues of dynamic call properties. Genetic variation in female
preferences widiin populations, which is necessary for genetic
correlations widi male traits to arise dirough indirect selec-
tion, has been demonstrated in a variety of animals (Bakker
and Pomiankowski, in press; reviews in Budin, 1994), but odi-
er studies have found diat such variation is meager or unde-
tectable (e.g., Budin, 1993).
METHODS
Recordings
Advertisement calls of H.
versicolor
were recorded widi high-
quality tape recorders (Nagra, Sony Pro-Walkman) and vari-
ous microphones (Sony, Sennheiser, Tandy); most recordings
were made between about
1
h after sunset and midnight; we
avoided making recordings after chorus activity began to
wane. After recording a male for die first time, we measured
his body size (snout-to-vent lengdi or mass) and marked him
for subsequent identification during die same breeding sea-
son, using numbered laundry tags attached to waist bands
made of embroidery floss or rubber bands. After every re-
cording, we measured die cloacal temperature of die frog or
die air temperature at die site.
A
subset of
males
was recorded
one or more additional times on different nights (mean =
3.1 nights; range: 2-15). Recordings were made at two ponds
diat were about 100 m apart in die Thomas Baskeu. Wildlife
Area in Boone County, Missouri, during four breeding sea-
sons:
1980, 1987, 1989 and 1990.
Acoustic and statistical analysis of call properties
We analyzed at least 10 calls per male per recording to obtain
estimates of CR, PN, and DC (=call duration divided by call
period, die time interval between die start of one call and die
start of die following call). Analysis of five calls per recording
was adequate to provide accurate estimates of
PR
and DF (see
Gerhardt, 1991). DF was estimated (±50 Hz) widi a Kay DSP
5500 Sona-Graph; all odier properties were analyzed with cus-
tom-designed software (G. Klump, D. Polete, and W. Cable)
and an IBM-compatible or NeXT personal computer. We did
not correct our measurements of CR, PN, DC, and DF for
temperature because correlations between die values of diese
acoustic properties and temperature were weak and usually
statistically nonsignificant (see also Gerhardt, 1991).
We computed die mean value and coefficient of variation
(CV = [standard deviation/mean] X 100) of each call prop-
erty from analyses of die first recording of each male; we used
diese data to estimate population means and patterns of
vari-
ability in each of die four
years.
Data from multiple recordings
of individual males (mean values from each recording) were
used to estimate repeatabilities of each property by
year,
using
die mediods oudined in Lessells and Boag (1987) and die
Systat statistical package (Wilkinson, 1989). We also estimated
die relative repeatability of the call properties as follows. First,
we computed die mean for each property on nights when we
recorded at least six males. Second, for each male present on
at least two such nights, we expressed die mean value of die
call properties of diat individual on each night in standard-
deviation units, i.e., mean value for die night of all recorded
males minus die individual's value divided by die standard
deviation of die property for die season. Third, we computed
die repeatability of these scores. These procedures allowed us
to quantify die extent to which some males consistendy pro-
duced calls widi values of pertinent call properties above or
below die nighdy mean values.
Playback experiments: general procedures
Syndietic advertisement calls were generated with a custom-
designed program (G. Klump and J. Brown), which created
eight-bit digital files widi an output sampling rate of 20 kHz.
The stimuli used to test females of H.
versicolor
had a PR of
20 pulses-sec"1, a DF of 2.2 kHz, and a second component of
1.1 kHz at6 dB; for tests of H.
chrysoscelis,
pulse rate was
50 pulses-sec"1,
DF
was 2.4 kHz, and diere was a second com-
ponent of 1.2 kHz at6 dB. The pulse shapes (amplitude-
time envelopes) were typical of die calls of males of die spe-
cies being tested (see Gerhardt, 1994a for sample oscillo-
grams). We combined pairs of alternative signals in a digital
file so diat each stimulus could be played back dirough a
separate channel (see below) widi a fixed timing relationship.
We adjusted diese relationships so diat equal intervals of si-
lence separated die end of one stimulus and die beginning
of die alternative stimulus (Figures 1 and 2a); dius, no stim-
ulus led or lagged its alternative. Pairs of alternative stimuli
differed in PN, CR, or bodi; values of diese properties were
widiin natural ranges of variation of die mean values observed
in typical populations of die two species. These stimuli lacked
die call-to-call variation typical of natural bouts, but experi-
ments widi H.
versicolor
indicated diat such variation does not
influence female preference if die mean values of alternative
stimuli are very similar (Gerhardt and Watson, 1995). Details
about die stimuli are given in die results section or in die
appropriate figures.
Most females (53% of H.
chrysoscelis
and 85% of H. versi-
color)
were tested in die temperature-regulated, semi-anechoic
chamber described in detail in Gerhardt (1994a). The re-
mainder were tested in a smaller chamber described in Ger-
hardt and Doherty (1988). The playback system for most ex-
periments consisted of
an
IBM-compatible personal computer,
digital-to-analog interface board (Supersound, engineering
by guest on July 6, 2011beheco.oxfordjournals.orgDownloaded from
Gcrhardt et al.
Dynamic call properties and mate choice in tree frogs
Figure 1
Oscillograms of synthetic calls used in female choice experiments
with females of
Hyla
versicolor.
The top trace shows one call with 24
pulses, and the lower trace two calls with 12 pulses each. The pulse
rate of all stimuli was 20 pulses-sec"1. The CR of the 12-pulse calls
was twice that of the 24-pulse call; thus, the total acoustic on-time
was the same for both alternatives. The timing relationship was
fixed so that neither alternative led nor lagged the other. About
one-third of the females first heard a 12-pulse call followed by a 24-
pulse call, another diird first heard the 24-pulse call, and the
remaining frogs first heard two 12-pulse calls. Similar stimuli were
used to test females of H.
chrysoscelis,
except that the pulse rate was
50 pulses-sec"1. Irregularities in the oscillograms are artifacts of
aliasing of the digital oscilloscope display.
version, Silicon Soft, San
Jose,
California), Nagra DSM ampli-
fiers and Analog-Digital-Systems 200 loudspeakers. A few ex-
periments employed the same speakers but used an analog
playback system consisting of a Denon DR cassette deck. We
checked the fidelity of the playback systems and acoustics of
the chambers at the points where females were released and
at many other positions within the sound fields produced by
the two speakers. The frequency response of the systems was
±2 dB over the range of interest, and the peak-to-peak am-
plitude of sound reflections was more than 35 dB less than
that of the direct sounds. Additional details about the exper-
imental setup appear in Gerhardt (1992, 1994a).
Females of both species of gray tree frogs were collected in
amplexus. Each was separated from her mate and taken to
the laboratory, where she was refrigerated for 1-3 nights to
inhibit oviposition.
We
tested most females within 24 h of
cap-
ture,
after acclimation for at least 30 min to the test temper-
ature of
20°C.
We later induced some of these frogs, and oth-
ers that were not tested within 3 days of capture, to respond
phonotactically by injections of progesterone and prostaglan-
din (protocol modified from Schmidt, 1985).
The playback levels of alternative stimuli were adjusted with
Bruel & Kjaer 2209, GenRad 1982 or 1933 sound level meters
to SPLs typical of those of calling males in the field (range:
84.5-95 dB "fast" RMS at 1 m; Gerhardt, 1975); maximum
SPL in playback experiments was 85 dB at the release point
and 91 dB at 1 m in front of a speaker. We placed females
individually in a round (8.5 cm in diameter, 3 cm in height)
hardware-cloth cage and restrained them until diey heard
three repetitions of each alternative stimulus; the first stimu-
lus heard was usually varied from female to female. We re-
leased the female from the cage by removing its Plexiglass top
remotely via a string. In the smaller chamber, we directly mon-
itored female movements on a grid (10-cm squares) from in-
side the chamber. In the larger chamber, we observed move-
ments from outside the chamber with a low-light television
camera (Sanyo CCD, Model VDC-2524) and monitor (Sanyo
VM 4130). A choice was recorded when a female moved to
within 10 cm of
a
speaker after having made movements char-
acteristic of phonotactic orientation, such as head and body
scanning during or immediately after sound emission. We not-
ed the time between the female's release and her choice; at
least 10 min separated tests of the same female.
Playback experiments: designs
Figure 2 shows arrangements of loudspeakers relative to the
release point of the female. Stimuli were switched between
speakers, and speaker positions were switched periodically to
control for potential directional biases or asymmetries in the
fidelity of the playback system; none were detected. Besides
changing the side on which the nearer speaker was placed in
the design shown in Figure 2d, we also changed the axis of
the setup in some tests, i.e., the release point was on one side
of the speakers in some tests and on the other side for other
tests.
Experiments also differed in the relative SPL and play-
back modes of the alternative stimuli as described in the sec-
tions that follow.
Experiments
1—5:
continuous playbackshigh-PN or fast-CR
calls versus Imu-PN or slow-CR calls with higher SPL at the
release
point
In Experiments 1 and 2, we released females of H.
versicolor
and H.
chrysoscelis
midway between two speakers separated by
2 m (Figure 2b). In Experiment 1, the SPL at the release point
of calls of high PN (24 pulses) was 79 dB and that of alter-
natives of
low
PN (12 pulses) was 85 dB. In Experiment
2,
the
SPL at the release point of high-CR calls (26.6 calls/min) was
79 dB and that of low-CR calls (13.3 calls/min) was 85 dB.
In Experiment 3, we released females of H.
versicolor
at a
point
1
m from one of two speakers diat were separated by 3
m (Figure 2c). Alternatives of 24 and 12 pulses had the same
broadcast amplitude (85 dB) at 1 m from the speaker; how-
ever, because the female was released 1 m from the source of
low-PN calls and 2 m from the source of high-PN calls, die
SPL at die release point of the low-PN and high-PN calls was
85 and 79 dB, respectively. In odier words, a 6-dB difference
favoring low-PN over high-PN calls
was
created by a difference
in distance from the release point to the sound sources rather
than by differential amplification. In Experiment 4, a differ-
ence of
6-dB
favoring slow-CR calls (13.3 calls/min) over high-
CR calls (26.6/min) was -created by distance as in Experi-
ment 3.
In Experiment 5, with females of H.
versicolor,
distance also
created a 6-dB difference in SPL favoring low-PN calls (12
pulses) over high-PN calls (24 pulses). However, die spatial
arrangement of speakers relative to the release point (Figure
2d) differed from that in Experiments 1-4; namely, die an-
gular separation between die speakers was about 36° radier
dian the usual angle of
180°.
Compared widi die usual design,
the spatial arrangement of Experiment 5 simulates more re-
alistically die geometry of calling males encountered by fe-
males in nature and should make die choice more difficult
because females moving toward die more distant speaker
would experience an even greater difference in SPL favoring
die closer, low-PN alternative dian diey did at die release
point. This increased difference, which occurs in an area be-
tween the release point and nearby speaker, is a simple con-
by guest on July 6, 2011beheco.oxfordjournals.orgDownloaded from
10Behavioral Ecology Vol. 7 No. 1
(a)(b)
-4dB J
+6dB
1
Release
point
o
79 dB or85dB
85 dB 85 dB
(C)
Playback of long Playback of long
calls resumed = calls terminated
decision point
-0.5 m-
(d)
Release
point
o
79
dB
35 dB
Figure 2
Diagrams of experimental setups for female choice experiments, (a) Cartoons representing oscillograms of pairs of stimuli that differed in
CR (two top traces) and PN (two lower traces), (b) Setup for Experiments 1-2 and 6-10. In experiments 1-2, the SPL of high-PN or fast-CR
calls from the shaded speaker (SPLs indicated by bold font) was adjusted to be 79 dB at the release point; the SPL of low-PN or slow-CR calls
from the unshaded speaker (SPLs indicated by italic font) was 85 dB. Alternative stimuli were played continuously until a female moved to
within 10 cm of a speaker. The same procedure was followed in Experiment 6, except that the SPL of high-PN, slow-CR calls and low-PN,
fast-CR calls were equalized at 85 dB at the release point. In Experiments 6-10, SPLs of alternative stimuli were also equalized at 85 dB at
the release point However, as the female moved to within about 0.5 m (dotted line) of the source of high-PN calls (shaded speaker),
playback of this stimulus terminated. When females moved back to a point about halfway between the original release point and the source
of low-PN calls (dashed line), playback of high-PN calls resumed. Experiment 11 used the same procedure; however, speakers were 4 m apart,
the SPLs of alternatives were equalized at 79 dB at the release point and thus were 75 dB and 85 dB at the decision point for the high-PN
and low-PN calls, respectively, (c) Setup for Experiments 3 and 4. The MB difference in SPL at the release point favored low-PN or slow-CR
calls (SPLs in italic font) over high-PN and fast-CR calls (SPLs in bold font), (d) Arrangement of speakers relative to the release point for
Experiment 5. High-PN calls played back from the shaded speaker, and the angular separation of the two speakers with respect to the release
point was about 36°. The SPL of each alternative was
91
dB at
1
m in front of the speaker from which it was broadcast, resulting in a 6-dB
difference at the release point favoring low-PN calls. The relative increases in SPL as the female moved toward the two speakers differed so
that, for example, at about 50 cm from the release point, the SPL of the high-PN calls had increased by about 2 dB while that of the nearby
low-PN calls had increased by 6 dB.
sequence of the fact that SPL increases by 6 dB as the distance
between source and receiver is halved.
Experiment 6: continuous playbacksequal call-duty
cycles
and
SPL
In Experiment 6 with females of both H.
versicolor
and //..
chrysoscelis,
the SPL of the two alternatives was 85 dB at the
release point midway between the speakers (Figure 2b). One
alternative had twice the PN and one-half die CR of die al-
ternative stimulus (Figure 1); thus, the DC or total acoustic
on-time of die alternatives was equal. The timing relationship
was fixed to yield equal periods of silence between the end of
die high-PN call and die beginning of die first low-PN call
and between die end of die second low-PN call and die be-
ginning of the high-PN caH.
Experiments 7-11: discontinuous playback of one alternative
sources equidistant from the
release
point and SPL of the
alternatives equal at the
release
point but not at the decision
point
In Experiments
7-11,
we released females of//, versicolor^ mid-
way between speakers, where die SPL of die two alternatives
was equal (Figure 2b). One speaker played back high-PN calls
and the other low-PN calls. The experiments differed in die
magnitude of die difference in die PN of die alternatives and
die distance between die speakers (2 m in Experiments 7-10
and 4 m in Experiment 11). Playbacks of high-PN calls were
terminated as die female moved toward die source of high-
PN calls and resumed after females turned around and moved
to about one-half of die distance between the original release
point and die source of low-PN calls. At diis decision point
(Figure 2b), females eidier continued toward die nearer
source of low-PN calls or reversed direction of movement and
returned to die source of high-PN calls.
Playback experiments: statistical analyses and interpretation
Our goal was to estimate die proportions of females diat
choose one of
two
alternative stimuli in die designs described
above. The 95%-confidence intervals for each observed pro-
portion provide a standard measure of die uncertainty about
diese estimates. More importandy, die lower bound on die
observed proportion may be a biologically significant estimate
even if diere is no statistically significant preference; e.g., in
by guest on July 6, 2011beheco.oxfordjournals.orgDownloaded from
Gerhardt
et
al.
Dynamic call properties
and
mate choice
in
tree frogs
11
Table 1
Descriptive statistics for four call properties of gray tree frog (Hyla venicolor) recorded at two ponds at the Baskett Research Area
Year
1980
1987
1989
1990
Property
Dominant frequency
(Hz)
Pulse number
Call duty cycle
Call rate (calls/min)
Dominant frequency
(Hz)
Pulse number
Call duty cycle
Call rate (calls/min)
Dominant frequency
(Hz)
Pulse number
Call duty cycle
Call rate (calls/min)
Dominant frequency
(Hz)
Pulse number
Call duty cycle
Call rate (calls/min)
A»
141
28
28
28
168
168
168
168
29
29
29
29
91
23
23
23
Mean
2132
16.7
0.210
13.5
2263
17.8
0.205
14.4
2264
15.4
0.168
12.7
2319
14.1
0.178
16.6
SD
189
4.4
0.036
4.5
90
4.3
0.043
2.5
129
3.4
0.038
3.8
155
3.1
0.040
5.6
CV
(%)
8.9
26.1
17.6
33.4
4.0
24.4
21.0
17.7
5.7
23.6
22.6
30.2
6.7
22.3
22.5
33.7
Maximum
difference
35
175
130
332
30
233
337
367
25
162
172
190
36
162
193
270
Repeatability
(N)b
0.40**
(28)
0.0 (28)
0.0 (28)
0.02
(28)
035***
(40)
0.37***
(40)
0.14*
(37)
0.21
(36)
0.62***
(29)
0.49***
(29)
0.14**
(29)
0.23**
(29)
0.77***
(23)
0.54***
(17)
0.49
(12)
0.23*
(17)
Number
of
males—estimates
of
mean
and
CV based
on
first recording
of
male.
b
Number
of
multiply recorded
males.
*
p
<
.05,
p
<
.01,
***
p
<
.001.
three experiments, the preference of females of H. venicolor
for less intense or more distant stimuli was not significant with
a binomial test, but the lower bound on the proportion of
females responding to these stimuli was 45% or higher (see
Figure 5). Furthermore, significance tests of the point null
hypothesis of no preference (using the usual alpha-level of
0.05) are heavily biased to avoid Type I statistical errors and
hence inevitably increase the chances of Type II errors, es-
pecially when sample sizes are relatively small. Gerhardt
(1992) provides complete discussion of interval estimation
(and its Bayesian interpretation) as applied to die results of
female choice tests. We nevertheless provide /^values from bi-
nomial and other significance tests for the convenience of
readers unfamiliar with or unaccepu'ng of our statistical ap-
proach.
RESULTS
Variation
in
advertisement calls
In Table I we present descriptive statistics for four properties
of the advertisement calls of males of H.
venicolor.
Mean val-
ues and between-male variation (CV, SD, and range of varia-
tion [maximum value minus minimum value expressed as a
percentage of die minimum value]) of die four properties
were similar from year to year. The data were consistent with
the pattern observed by Gerhardt (1991) for die 1987 season
(values for die CV and ranges of variation are taken from diat
paper, estimates of repeatability are updates based on analyses
of additional males): overall between-male variation in die dy-
namic properties (PN, CR, and DC) was much greater dian
diat of die static property, DF. DF was die only variable diat
was repeatable during all 4 years, but each dynamic property
was repeatable during at least 2 years. The repeatability for
PN was comparable to diat for DF in 3 of die 4 years. For
every call property, some males produced calls on two or more
nights widi mean values diat differed by about one standard-
deviation or more from die grand mean based on all males
recorded on die same nights (Figures 3 and 4).
In Table 2 we summarize die between-male variability on
32 nights when six or more males were recorded. These nights
include diose when die males whose calls were analyzed in
Figures 3 and 4 were recorded plus 6 additional nights in
1980.
The average standard deviation and range of variation
of die four call properties observed during diese nights were
similar to die seasonal standard deviations and ranges of vari-
ation for 1987 and 1990 (Table 1). Thus, we used die seasonal
standard deviations to calculate die standard-deviation units
diat are plotted in Figures 3 and 4.
Female preferences
Fifty of 59 females (85%) of H.
chrysoscelis
from six localities
in Missouri (56 of die females were from Laclede or Phelps
County) responded in at least one experiment in 1990-92.
We observed responses in 81 of
91
tests; diese totals include
only one test per female per experiment
Eighty-nine percent of
207
females of H.
venicolor
from six
localities in Missouri (192 of die females were from Boone or
Cole County) responded in at least one experiment in 1990-
94.
We observed responses in 296 of 347 tests, including sec-
ond responses of 84 females tested twice.
We induced phonotaxis in 28 of die 59 females of H. chry-
soscelis
and 20 of die 207 females of H.
venicolor
using pro-
gesterone-prostaglandin; diese females did not differ from
those tested widiin a few days of capture in die proportions
diat responded or in dieir choices.
Experiments 1-5: continuous playbackshigh-PN or fast-CR
calls versus low-PN or slow-CR calls with higher SPL at the
release
point
The majority (S67%) of die females of bodi species chose
high-PN calls over low-PN calls (Experiment 1) and high-CR
calls over slow-CR calls (Experiment 2), even diough the SPL
of die preferred stimuli was adjusted to be 6 dB less at die
release point dian diat of die alternatives (Figure 5). The
preference for high-PN calls in H.
chrysoscelis
was especially
strong.
Only females of H.
venicolor
were tested in Experiments 3
and 4, in which die 6-dB difference in SPL at die release point
was created by distance (Figures 2c and 5). Whereas about
70%
of die females of H.
venicolor
chose die source of fast-
CR calls (Experiment 4), all ten females chose die more dis-
by guest on July 6, 2011beheco.oxfordjournals.orgDownloaded from
12Behavioral Ecology Vol. 7 No. 1
10 20 30 40
Ranked frog number
5
2
a
S-i
TH
« -2
r.= 0
-11
5 10 15
Ranked frog number
20)
Figure 3
Relative repeatabilities of four
call properties in 1987. Each
point represents the mean val-
ue of an individual's calls dur-
ing one night in standard de-
viation units of the mean value
(shown as the dotted horizon-
tal line labeled "zero") based
on values from at least six
males recorded on the same
night. Males are ranked in or-
der of increasing average value
of their means. The number of
points per male indicate the
number of nights on which his
calls were recorded; some val-
ues were so similar that they
appear to be one point The
values of
r,
are estimates of rel-
ative repeatability. See the text
for additional details.
a
w 3
H*
a
9
« 2
a1
0
4J
JJ
0
5-2
PS-1
r
-
.
...
I
-
1
. = 0
I
T
I 1
it IT1
11
-
-
5 10 15
Ranked frog number
205 10 15
Ranked frog number20
tant source of high-PN calls over Iow-PN calls played back
from the nearer source (Experiment 3).
All ten females of H.
versicolor
chose high-PN calls broad-
cast from the distant speaker when the two speakers were sep-
arated by 180° (Experiment 3); a smaller proportion (0.72)
of females first chose high-PN calls broadcast from the distant
speaker when the two speakers were separated by 36° (Exper-
iment
5;
Figures 2d, 5, and 6a; G = 6.0, p < .05). Considering
both responses of each female, the proportions of frogs in
each of the four possible response categories were close to the
proportions expected under the hypothesis that each female
had a fixed probability of .72 of choosing the high-PN stim-
ulus (Figure 6b).
Experiment 6: continuous playbacksequal coil-duty
cycles
and
SPL
In experiment 6, about 85% of the females of H.
chrysoscelis
and about 72% of the females of H.
versicolor
initially chose
high-PN calls (24 pulses) emitted at a slow CR rather than
low-PN calls (12 pulses) played back at a fast CR (Figure 7a).
Considering both responses of the 31 females of H.
versicolor
that responded twice in this experiment, the proportions of
frogs in each of the four possible response categories were
close to the proportions expected under the hypothesis that
each female had a fixed probability of .72 of choosing the
high-PN, slow-CR stimulus (Figure 7b).
Experiments 7-11: discontinuous playback of one alternative
sources equidistant from the
release
point and SPL of the
alternatives equal at the
release
point but not at the decision
point
In Experiments 7-10 with females of H.
versicolor,
the pro-
portions of females that reversed direction, moved away from
a low-PN call, and returned to a source of high-PN calls de-
pended on the difference in PN between the alternatives and
on whether the PN of the low-PN call was at the low end of
the range of variation in the Missouri population (i.e., 12 puls-
es;
see Table 1). In Experiment 7, about 67% of the females
returned to a source of 18-pulse calls rather than continuing
to a nearby source of 12-pulse calls (Figure 8); a PN of 18 is
close to the mean in 1987, the season for which we had the
largest sample size (see also Gerhardt, 1991). About 75% of
the females returned to the source of high-PN calls when the
high-PN call had 24 pulses and the low-PN alternative had 12
pulses, a difference of 100% in PN (Experiment 8: Figure 8).
Most females continued to the source of a low-PN call with 18
pulses rather than returning to the source of high-PN calls
with 24 pulses (Experiment 9) or 30 pulses (Experiment 10).
When the distance between the speakers
was
4 m (Experiment
11),
females usually continued to the source of low-PN calls,
even though the two alternatives had 12 and 24 pulses, re-
spectively (Figure 8).
by guest on July 6, 2011beheco.oxfordjournals.orgDownloaded from
Gerhardt et al.
Dynamic call properties and mate choice in tree frogs
5 10 15 20
Ranked frog number
255 10 15 20 25
Ranked frog number
5 10 15 20 25
Ranked frog number
5 10 15 20
Ranked frog number25
Figure 4
Relative repeatabilities of four
call properties in 1989. See the
legend of Figure 3 and the text
for further information.
DISCUSSION
Seasonal variability in call properties
Significant within-season variation among males in call prop-
erties is a prerequisite for an evolutionary response to selec-
tion and for indirect selection on female preferences. As dis-
Table 2
Variability of advertisement calls within choruses on 32 nights*
Average range
of variation
(range of
Variable Mean SD ranges, %)b
Dominant frequency
Pulse number
Call duty cycle
Call rate
129 Hz
2.8 pulses
0.03
3.2 calls/min
16 (6-30)
83 (23-144)
89 (30-328)
118 (41-373')
Mean of 9.7 males (range 6-23) were recorded on each night (6
nights in 1980, 16 nights in 1987, and 10 nights in 1989).
b Ranges of variation are expressed as a percentage of the minimum
value.
c The maximum range for call rate exceeded the 1987-value for the
season reported in Table 1 because only the first recording of each
male was considered, whereas the data presented here included
values from additional recordings of the same males.
cussed in Gerhardt (1991), dynamic properties, such as PN,
CR and DC, were more variable within bouts of calling than
static properties such as DF; the between-male variance (CV)
of dynamic properties was also greater than that of
DF.
Nev-
ertheless, during one or more breeding seasons, the repeat-
ability of dynamic properties was comparable to that of static
properties. Evidence for statistically significant, within-season
variability among males has been reported in other studies of
anurans {H.
versicolor:
Gerhardt, 1991; Runkle et al., 1994;
Sullivan and Hinshaw, 1992; H.
arborea.
Friedl, 1992; Acris
crepitans.
Wagner, 1989;
Pseudacris
crudfer:
Sullivan and Hin-
shaw, 1990; Bufo
valiiceps:
Wagner and Sullivan, 1995). Com-
parable data from insects and birds are reviewed by Boake
(1989) and Gerhardt (1991).
Repeatability and other measures of seasonal among-male
variation are a first step in assessing the possibility that partic-
ular male properties will respond evolutionariry to selection.
Repeatability sets an upper bound on heritability (Boake,
1989);
highly repeatable traits vary little within individuals and
considerably among individuals. However, highly repeatable
traits are not necessarily heritable. Moreover, estimates of
both repeatability and heritability are likeh/ to vary among
seasons and different studies because of environmental vari-
ation and different sampling procedures (see also Aspi and
Hoikkala, 1993).
In this study, estimates of repeatability varied from year to
year; estimates of all three dynamic properties were near zero
in 1980, whereas the same properties showed significant re-
by guest on July 6, 2011beheco.oxfordjournals.orgDownloaded from
14
Behavioral Ecology Vol. 7 No. 1
Exp.
1
Exp.
2
Exp.
3
Exp.
4
HC
24/
26
HV
20/
30
HC
22/
30
HV
21/
31
H-PN
F-CR
-6dB -6dB
By amplifier
H-PN
F-CR
-6 dB -6
dB
By distance
Figure 5
Percentage of female choices (one per female; HV = H.
versicolor,
HC •= H.
chrysoscelis)
for high-PN (H-PN) or fast-CR (F-CR) calls
over low-PN (I^PN) or slowOt (S-CR) calls. In Experiments 1 and
2,
a difference in SPL of 6 dB at the release point in favor of low-
PN or slow-CR calls relative to high-PN and fast-CR calls was created
by differential amplification; females were released midway between
the speakers as in Figure 2b. In Experiments 3 and 4, the 6-dB
difference was created by distance as shown in Figure 2c. Error bars
represent 95% confidence limits on the proportions of females that
responded to the preferred stimulus; raw data are also presented in
each bar as the number of females choosing the preferred stimulus
divided by the number of responding females, p-values for two-
tailed binomial tests of the null hypothesis of no preference:
Experiment 1—HC (<.001); HV (.099); Experiment 2—HC (.016);
HV (.07); Experiment 3—HV (<.01); Experiment 4—HV (.12).
peatability in 1989. Sullivan
and
Hinshaw (1992) also reported
between-year variation
in
repeatability estimates
for
call prop-
erties
in H.
versicolor.
Estimates
of
maximum repeatability
for
PN
and
DC were similar
in
Maine
and
Missouri (Sullivan
and
Hinshaw used equivalent measures
of
these properties, call
duration
and
calling effort, respectively).
The
highest repeat-
ability
for CR in the
Maine population
was
about twice that
for
the
Missouri population
(0.50
versus 0.23). Estimates
for
the static property
(DF)
were also comparable between Maine
and Missouri (maximum values
of
0.65
and
0.77, respectively).
In
a
one-year study
of
H.
versicolor
in Connecticut,
the val-
ues
of
dynamic properties
of
individual males were relatively
consistent throughout most
of the
evening; males tended
to
reduce
PN
late
in the
nightly chorusing period
and
toward
the
end of the
season
as
distances between calling males
in-
creased (Runkle
et al.,
1994). Moreover, males that call
in
dense aggregations tend
to
produce calls with higher
PNs at
slower
CRs
than more isolated males (Runkle
et al., 1994;
Wells
and
Taigen, 1986). Thus,
our
estimates
of
repeatability,
as those
of
other studies,
are
subject
to
errors because,
al-
though we avoided recording males after
the
chorus began
to
diminish,
we did not
always record males
at die
same time
during
the
early hours
of
each chorus
and
because we includ-
ed data from some recordings
if a
male
was
recorded early
on
one
evening
and
late
on
another evening,
or
early
and
late
in
the
season; between-male variability
on a
given night could
be overestimated
if
some males were recorded early
and oth-
ers late
or if the
spacing
of
recorded males differed signifi-
cantly.
Besides
the
limitations discussed above, conventional esti-
mates
of
repeatability
are
probably
not the
best measures
for
estimating
the
potential
for
selection
on
individual males over
the course
of a
season. Individual females usually assess
and
mate with
a
male during
a few
hours
on a
single night, where-
as repeatability
is
typically estimated from recordings
of dif-
ferent males made over
the
course
of
a
season. Aldiough some
property
of a
male's call might
be
highly consistent between
nights,
he
would have higher than average mating success
only
if the
value
of die
property were more attractive than
that
of
most other males
on
most
of the
nights
he
attended
the chorus.
The
relative repeatability measure presented here
takes diis pattern
of
choice into account
(see
Wagner
and
Sullivan, 1995,
for a
similar approach),
and the
graphical view
of
die
seasonal data (Figures
3 and 4)
show
the
pattern
of
seasonal phenotypic variation
in
calls upon which female
choice potentially operates. These graphs make clear that
when relative repeatability was high,
the
calling performance
(PN,
CR, DC) or DF of
more males was consistently above
or
below
the
nightly means than when relative repeatability
was
low. However,
no
statistical
or
graphical analysis alone
can pre-
Flgure 6
Results of Experiment 5 with
females of H.
versicolor.
The
setup is diagrammed in Figure
2d. (a) Percentage of first re-
sponses to the high-PN (H-PN)
calls played back from the
more distant loudspeaker;
95%-confidence limits and raw
data are presented as in Figure
5 (/walue for two-tailed bino-
mial test <.001). (b) Percent-
ages of the 53 females that
made each of the four combi-
nations of first and second
choices appear as clear bars.
The chi-square value is for a
goodness-of-fit test of the hy-
pothesis discussed in the text,
and the expected binomial
percentages of females in each
category appear as solid bars.
Observed and expected num-
bers of females in each cate-
gory are also given.
(a)
100
90
g
80
I70
0.
o
6
60
50
40
30
20
10
38/53
First responses
(b)
=0.79,
dfe1,P=0.37
L-PN
L-PN1st response
2nd response
by guest on July 6, 2011beheco.oxfordjournals.orgDownloaded from
Gerhardt et al.
Dynamic call properties and mate choice in tree frogs IS
(a)
8
r
s
to
z"
1-P
&
si
B
CO
.8
0
100
90
80
70
60
50
40
30
20
10
0
-
. HC
-
-
-
-
-
-
1
30/35
HV
1
1
28/3S
First responses
^
CD
twic
ing
£
5
CO
CD
.9
0
O
100
90
80
70
60
50
40
30
20
10
-
-
-
-
m
H-PN,
S-CR
H-PN,
S-CR
HV
X2=0.94, c
rm
i
m
H-PN,
S-CR
L-PN,
F-CR
#=1,
P=0.33
n
i
m
L-PN,
F-CR
H-PN,
S-CR
JL.2.5
m
L-PN,
1st
response
F-CR
L-PN,
2nd
response
F-CR
Figure 7
Results of Experiment 6. The
setup is diagrammed in Figure
2b and the procedure ex-
plained in the legend of Figure
2.
(a) Percentages of first re-
sponses by females of H.
verri-
color
(HV) and H.
chrysoscdis
(HC) to the high-PN, slow CR
calls;
95%-confidence limits
and raw data are presented as
in Figure 5 (Rvalues for two-
tailed binomial test were <.001
for HC and <.01 for
HV).
(b)
Percentages of the 31 females
that made each of the four
combinations of first and sec-
ond choices appear as clear
bars.
The chi-square value is
for a goodness-of-fit test of the
hypothesis discussed in the
text and the expected percent-
ages of females in each cate-
gory appear as solid bars. Ob-
served and expected numbers
of females in each category are
also given.
diet the magnitude of female choice simply because statistical
discrimination is not equivalent to perceptual discrimination.
Nightly variation in dynamic properties and dominant
frequency: the scope for female choice
The individual female samples some subset of males that are
at the chorus on one or two nights during a season when she
mates.
Information about minimum differences in calls that
elicit female preferences is valuable for predicting the relative
mating success of males. For example, the minimum mean
PN of five of the 23 males recorded multiple times in 1989
was a standard deviation, or 3 pulses/call (Table 1), higher
than the grand nightly means (males 18-23 in Figure 4). This
same information allows us to assess the scope for female
100
g 90
^ 80
S 70
a.
60
I 50
O)
Z 40
1
o
30
20
O
10
0
Exp.
7
Exp.
S
Exp.
11
Exp.
9
Exp.
10
2O3121/28 6/23V20
12pvs
18p
12pvs
24p
18pvs
24p
18pvs
30p
Figure 8
Results of Experiments 7-11 with females of//,
versicolor
in
the
setup diagrammed in Figure 2b. The bars show the percentages of
females that moved from the decision point back toward the source
of more distant, high-PN calls. In Experiments 7-10, speakers were
2 m apart, and in Experiment 11 speakers were 4 m apart.
Confidence limits and raw data are presented as in Figure 5. p-
values for two-tailed binomial test; Experiment 7—.14; Experiment
8—.13;
Experiment 9—.012; Experiment
10—.041;
Experiment
11—.11.
choice on a nightly basis; such choice could have evolutionary
effects on female preferences even if differences in male calls
are mainly determined environmentally, e.g., if fertilization
ability were determined by feeding success and indicated by
call variation. To the extent the available data permit, we con-
sider both minimum differences in a property that result in
preferences when alternative stimuli have the same SPL and
minimum differences that result in SPL (distance)-indepen-
dent preferences.
Females should have numerous opportunities to express
preferences based on PN. Experiments with H.
versicolor
in-
dicate diat females preferred calls with just two more pulses
per call than alternatives (PNs of 10 versus 8 and 20 versus
18,
differences of 25% and 11%, respectively; Gerhardt and
Watson, 1995; H. Gerhardt, S. Tanner, and M. Proett, unpub-
lished data). In the experiments reported in this article, we
found that when the difference in PN was 50%, about 67%
of the females of//,
versicolor
returned to the source of high-
er-PN calls that were 10 dB less than those of the lower-PN
alternative at the female's decision point (Experiment 7: 12
versus 18 pulses per call). When the difference in PN was
100%,
the proportion of females making such a choice in-
creased to 75% (Experiment 8: 12 versus 24 pulses per call).
Similarly, when the difference in PN was 100%, most females
of both species chose the high-PN stimulus with an SPL that
was 6 dB less than that of the low-PN alternative (Experiment
1);
all females of H.
versicolor
chose the high-PN call when
the 6 dB difference in SPL
was
created by distance rather than
differential amplification (Experiment 3). On each of the 32
nights when six or more males were recorded (Table 2), at
least one pair of males produced calls that differed in mean
PN by at least 3 pulses/call. Between-male differences in mean
PN of
50%
or greater were observed on 24 of 32 nights, and
differences of 100% or more were observed on 10 nights.
Moreover, even though males tend to match the calling per-
formance of their neighbors (Runkle et al., 1994), 11 of 23
pairs of males that we recorded within 15 min and 2 m of
each other produced calls that differed by at least 3 pulses/
call (grand mean of the 23 pairs: 3.1; range : 0-13; Table 3).
Observed differences in CR were also sufficient for females
to exercise choice on many nights. We do not have estimates
by guest on July 6, 2011beheco.oxfordjournals.orgDownloaded from
16
Behavioral Ecology Vol. 7 No. 1
Table 3
Summary of observations of mating choices of females of H.
venicolor in natural choruses
Distance from position of
Time to
mating
(min)
10
15
12
12
25
12
4
7
<10
<15
<15
and male (m)
Male
0.9
4.0
2.0
1.0
0.8
0.5
2.0
2.0
4.0
1.0
1.0
o
Next closest
male
2.0
>6.0
>6.0
2.0
1.0
2.0
2.0*
2.0
1.0
0.5
0.5
Angle between two
-
males at position
of female
(degrees)
180
180
not noted
not noted
30
not noted
o-
60
30
not noted
60
Unsuccessful male in tree was 0.75 m above successful male.
of the minimum difference
in
CR needed
to
elicit preferences
at equal SPL,
but in
Experiments
2 and 4,
when there was
a
difference
in CR of
100%, about two-thirds
of the
females
chose high-CR stimuli
6 dB
less than low-CR calls
at the re-
lease point.
We
observed between-male differences
of
100%
or more
in
mean
CR on 17 of
32 nights
(see
also Table
2).
Although estimates
of
repeatability
in
DF were more often
higher than those
of
the dynamic call properties,
our
analyses
suggest that nighdy variation
in
this property provided litde
scope
for
female choice. Females
of
H.
versicolor
preferred
a
call with
a
DF close
to the
mean
in the
population
(2.2 kHz)
over alternatives
of
lower